Maybe this is fixed in 2.0.0? See https://issues.apache.org/jira/browse/KYLIN-2135
Is this the same issue? It should split dims to higher reducer it seems... Although we are on 1.6. If this will fix our issue - can i apply this patch? I may look at what is all there vs 1.6. Any ideas? On Wed, Mar 14, 2018 at 1:44 PM, Alberto Ramón <[email protected]> wrote: > You can monitoring your yarn in step 3 > In any case, step 3 is a sample of Fat table to estimate number of keys > for each dim > If this step takes a lot of time, you will need review your cube design > > Alb > > On 14 March 2018 at 16:54, Sonny Heer <[email protected]> wrote: > >> 8 YARN nodes with 11 slots each. each slot is configured to ~2gb. Step >> #3 in Kylin is launching 19 mappers and 5 reducers. 5 reducers when there >> are 88 slots. >> >> btw: kylin version is 1.6 >> >> On Wed, Mar 14, 2018 at 9:48 AM, Sonny Heer <[email protected]> wrote: >> >>> YARN is properly configured. we use many other m/r and spark programs >>> that utilize the full slots. It's only when building cubes. >>> >>> On Wed, Mar 14, 2018 at 9:46 AM, Alberto Ramón < >>> [email protected]> wrote: >>> >>>> You need check your yarn configuration first >>>> >>>> On Wed, 14 Mar 2018, 14:58 Sonny Heer, <[email protected]> wrote: >>>> >>>>> Step 3 isn't using our full cluster. How can i increase the >>>>> mappers/reducers to use all the slots? Any config to look at in kylin? >>>>> >>>>> Thanks >>>>> >>>> >>> >> >
